Details
Pharmacoepidemiologic studies based on administrative claims data are often limited in their ability to control confounding, either due to unavailability of certain clinical information or questionable accuracy and/or completeness of coded information. Some variables (e.g., smoking, obesity) used in post-marketing observational studies are documented inconsistently or rarely because they do not impact a health plan’s decision for reimbursement. Hence, studies that use such variables to adjust for baseline risk imbalance may result in biased estimates for the association between medical products and health outcomes. The ability to capture these variables in administrative claims data may be improving over time, but little empirical data are available to support this statement. FDA conducted this study in the Sentinel Distributed Database (SDD) to assess the evolution of well-known confounding conditions (obesity, overweight, smoking, alcohol abuse or dependence, drug abuse or dependence, history of Coronary Artery Bypass Grafting or Percutaneous Transluminal Coronary Angioplasty, history of sudden cardiac arrest) that may be incompletely captured in U.S.-based healthcare claims databases. Results indicated a continuous increase in the recording of these confounding conditions in claims data, though low prevalence suggests that the conditions remain inadequately documented in U.S. claims. This exploratory Sentinel study provided information to FDA on the availability of common confounders in the SDD to better aid decisions on the use of Sentinel data and additional or alternate data sources for capturing these confounders.